Nonlinear adaptive filter design for integrated vehicle handling dynamics state estimation

This thesis considers nonlinear filter design for integrated vehicle handling dynamics state estimation. Such a state estimator is needed as not all of the vehicle states can be measured directly by the existing sensors, mostly due to reliability and economical reasons. Accurate information about vehicle handling states is essential for vehicle chassis control and chassis design evaluation. This study considers mathematical model-based filtering methods. A nonlinear 6DoF vehicle model employing an intermediate tyre magic formula is developed for the filter basis. The main problem faced by such a model-based filter is model uncertainties, especially in tyre parameters. The main objective of this study is to design filters which are robust against model uncertainties. Two nonlinear filtering methods are investigated: extended Kalman filter (EKF) and nonlinear robust filter (NRF). The EKF relies on accurate nominal model and ideal white/time uncorrelated assumption about model error noises. In contrast, the NRF tolerates inaccuracy of the nominal model as it accounts for the time-correlated behaviour of the model errors more properly. [Continues.]